Effective date: 01/01/2026 Last updated: 12/04/2026

Saudi ICT Shopper News uses artificial-intelligence (AI) tools as part of producing the articles you read. We believe readers have a right to know how the news they consume is made. This page describes β€” in plain language and with concrete detail β€” what AI does on this site, what it does not do, and who is accountable.

This disclosure is referenced from our Editorial Standards, our Corrections Policy, and our Privacy Policy.

1. The short version

  • We use AI to assist with drafting, translating, image generation, and SEO.
  • Every article is reviewed by a human editor before publication.
  • AI does not decide what is newsworthy, what is true, or what gets published.
  • When AI gets something wrong, the editor is accountable, not the model.
  • We do not allow AI models to be trained on our published articles without permission (see Section 6).

2. What AI does in our workflow

When a press release, document, or URL enters our publishing system, the typical AI-assisted pipeline runs in five stages:

Stage What AI does What the human editor does
1. Extraction Pulls text from URLs, PDFs, DOCX, scraped pages Decides whether the source is worth covering
2. Drafting Produces an English and Arabic draft from the source material under our editorial prompt Reviews accuracy, tone, framing, headline; rewrites where needed
3. Visual Generates a lead image and may auto-resize uploaded photos Approves the image, replaces it, or chooses a different one
4. SEO Suggests headline, slug, meta description, categories, tags Approves or overrides each field
5. Publishing Pushes to WordPress and our social channels per editor schedule Edits any post-publication issue and triggers the Corrections Policy when needed

A human editor signs off on stages 2-5. Nothing is published unattended.

3. AI models we use

We use multiple AI providers, chosen per task for quality, language coverage, and cost. As of the “Last updated” date above:

Provider Models Used for
Anthropic Claude Opus, Claude Sonnet Primary drafting, editorial review, translation
Google Gemini 3 Pro, Gemini Flash Drafting fallback, image generation, image analysis
OpenAI GPT-4.1, GPT Image Drafting fallback, image generation
Z.AI GLM Drafting fallback
xAI Grok, Grok Imagine Drafting fallback, image generation
DeepSeek DeepSeek Drafting fallback

The exact model used for a given article is logged internally and can be disclosed on request. We change models as the field evolves and update this page when we do.

We use these models through their paid APIs. Under the providers’ current terms (as of the “Last updated” date), content submitted via paid APIs is not used to train their models. If a provider changes that policy, we will either negotiate an exception, switch provider, or update this page.

4. What AI does not do

To be explicit:

  • AI does not select what stories we cover. A human editor decides that.
  • AI does not declare a claim “true” or “false”. Human editors fact-check.
  • AI does not write opinion or editorial-position pieces. Those are human-only.
  • AI does not name anonymous sources, fabricate quotes, or impersonate people.
  • AI does not process payments, contracts, or anything legally binding on the publication.
  • AI does not write articles about a specific identifiable individual based on protected-category attributes alone.

If we add a new category of AI use, we will update this page before the change ships.

5. Labeling

We do not label every article as “AI-assisted” because virtually every article on this site has some AI in the pipeline (translation, image resize, SEO). A blanket label would be noise.

We do label specifically when:

  • An image was generated by AI and depicts a real person, real event, or could be mistaken for documentary photography. Label: “AI-generated illustration” (or Arabic equivalent), shown in the image caption.
  • An entire article is published verbatim from an external press release without our own reporting. Label: “Press release” or “Industry News”.
  • A translation has been machine-translated without a fluent human editor pass. We currently do not publish articles in this category β€” translation always goes through a human editor.

These labeling rules are evaluated in every article during editorial review.

6. AI training restrictions

The articles, images, and underlying code on this site are protected by copyright. They may not be used to train, fine-tune, or evaluate AI models without our prior written permission.

We enforce this through several signals that automated crawlers are expected to respect:

  • robots.txt explicitly disallows known AI crawlers (e.g. GPTBot, ClaudeBot, CCBot, anthropic-ai, Google-Extended, PerplexityBot, Bytespider, Amazonbot, Applebot-Extended).
  • IETF AI preferences β€” we set X-Robots-Tag: noai, noimageai on HTTP responses where appropriate.
  • HTML <meta> β€” <meta name="robots" content="noai, noimageai"> where appropriate.
  • Terms of Use Section 8 explicitly forbids use of our content for AI training without permission.

For commercial licensing of our content for AI training: [email protected].

7. Personal data and AI

When we process a press release that contains personal data (an executive’s name, a contact’s email), that data flows through AI APIs as part of drafting. We minimize what we send: we do not include data unrelated to the editorial purpose, and we do not send any data classified as sensitive under the Saudi Personal Data Protection Law (PDPL) unless it is already public and central to the news.

Full data-flow detail is in our Privacy Policy Section 4.

8. When AI gets it wrong

AI systems make mistakes. They invent statistics, mis-name organizations, misattribute quotes, and sometimes invent entire facts (“hallucinations”). We treat AI errors the same way we treat human errors: a correction is issued and the editor responsible is identified.

For systemic AI errors β€” a pattern affecting multiple articles β€” we:

  1. Audit the affected articles and correct them
  2. Update our editorial prompts to prevent recurrence
  3. Add a note to our annual editorial review describing what changed

We do not blame the model. If a story has our byline, we are accountable for it.

9. Transparency reports

In our annual editorial review we publish, at minimum:

  • The set of AI models in use that year
  • The volume of articles processed
  • The number of AI-related corrections issued
  • Any change to the AI pipeline that materially affects readers

10. Reader feedback

If you spot an AI-generated error, or if you have a comment on how we use AI, write to [email protected] with the article URL and the issue. AI-related feedback is treated the same as any other editorial feedback.

11. Updates

This page is updated when our AI pipeline changes materially. Material changes include adding or removing a provider, expanding what AI is used for, or changing the human-review process. The “Last updated” date at the top reflects the current version.

12. Contact

Editorial: [email protected] Corrections: [email protected] Licensing requests for AI training: [email protected]